Apparatus for adjusting focus of camera and control method thereof
Disclosed are an apparatus and method for adjusting the focus of a camera. The apparatus in accordance with one embodiment of the present disclosure includes an image acquirer; a neuromorphic hardware including a plurality of neurons; and a controller configured to generate input vectors by vectorizing image data, input each of the generated input vectors to the neuromorphic hardware, select neurons, receive an edge pattern type value between the input vector and one corresponding edge pattern from each of the selected neurons, calculate a focus parameter value of the image frame on the basis of the received edge pattern type values, compare the focus parameter value of a current image frame and a focus parameter value of a previous image frame, determine a moving direction of the focus lens according to a comparison result, and move the focus lens on the basis of the determined moving direction.
1. An apparatus for adjusting the focus of a camera which automatically adjusts a focus by moving a focus lens through a focus lens driving motor, the apparatus comprising:
an image acquirer configured to obtain an image frame;
a neuromorphic hardware including a plurality of neurons connected through a parallel bus, wherein the neurons store edge pattern vectors generated by vectorizing previously learned edge patterns by which edges of an image are specifiable; and
a controller configured to set a specific region as a region of interest (ROI) on the basis of a center of the image frame input from the image acquirer, segment the set ROI into units of predetermined windows, generate input vectors by vectorizing image data obtained from each of the windows by sliding the windows, input each of the generated input vectors to the neuromorphic hardware, select neurons, each having an edge pattern vector that is most similar to each of the input vectors from the plurality of neurons, receive an edge pattern type value and a relative distance value between the input vector and one corresponding edge pattern from each of the selected neurons, calculate a focus parameter value of the image frame on the basis of the received edge pattern type values and the relative distance values, compare the focus parameter value of a current image frame and a focus parameter value of a previous image frame, determine a moving direction of the focus lens so that a focus with respect to the image frames increases according to a comparison result, and move the focus lens through the focus lens driving motor on the basis of the determined moving direction.
2. The apparatus of claim 1 , wherein the controller calculates the number of edge patterns of the image frame and an average of the relative distance values on the basis of the received edge pattern type values and relative distance values, and calculates the focus parameter value of the image frame on the basis of the calculated number of edge patterns and average value.
3. The apparatus of claim 2 , wherein the controller calculates the focus parameter value (F) through the following Equations 1 and 2:
F
m
=
N
F
+
μ
1
D
F
Equation
1
D
F
=
1
N
F
∑
i
=
1
N
F
dist
i
,
Equation
2
where F m represents a focus parameter calculation value of an m th frame, N F represents the number of windows recognized as edges by the neuromorphic hardware, D F represents an average of distance values of the windows recognized as edges, dist i represents a distance value of one window recognized as an edge, and u represents a coefficient for adjusting a ratio between two parameters having different physical characteristics.
4. The apparatus of claim 1 , wherein image edge pattern vectors stored in the plurality of neurons of the neuromorphic hardware are one of vertical, horizontal, 45-degree diagonal, and 135-degree diagonal edge patterns.
5. A method of controlling an apparatus for adjusting the focus of a camera which includes an image acquirer configured to acquire an image frame, a neuromorphic hardware configured to include a plurality of neurons connected through a parallel bus, wherein the neurons store edge pattern vectors generated by vectorizing previously learned edge patterns by which edges of an image is specifiable, a focus lens driving motor configured to move a focus lens, and a controller, the method comprising:
setting a specific region as a region of interest (ROI) on the basis of a center of the image frame input from the image acquirer;
segmenting the set ROI into units of predetermined windows;
generating input vectors by vectorizing image data obtained from each of the windows by sliding the windows;
inputting each of the generated input vectors to the neuromorphic hardware;
selecting neurons, each having an edge pattern vector that is most similar to each of the input vectors from the plurality of neurons;
receiving an edge pattern type value and a relative distance value between the input vector and one corresponding edge pattern from each of the selected neurons;
calculating a focus parameter value of the image frame on the basis of the received edge pattern type values and the relative distance values;
comparing the focus parameter value of a current image frame and a focus parameter value of a previous image frame;
determining a moving direction of the focus lens so that a focus with respect to the image frames increases according to a comparison result; and
moving the focus lens through the focus lens driving motor on the basis of the determined moving direction.
6. The method of claim 5 , wherein the calculating of the focus parameter value includes calculating the number of edge patterns of the image frame and an average of the relative distance values on the basis of the received edge pattern type values and the relative distance values, and the calculated focus parameter value of the image frame increases as the calculated number of edge patterns increases and the average decreases.